Classification of estrogen receptor-β ligands on the basis of their binding affinities using support vector machine and linear discriminant analysis

Pharmaceutical Perspectives of Nonlinear QSAR Strategies

“…88 Dong et al obtained a percentage (%) of correct predictions of 93% and 82.6% on the training and test set, respectively. A well performing SVM classification model has been carried out also for estrogen receptor-β ligands (105 diphenolic azoles) by Luan et al 89 In this case, the accuracy of the model prediction on the test set, comprising 26 compounds, was 91.4%, a percentage (%) value higher than the accuracy achieved by the linear discriminant analysis (LDA), based on the same descriptor set. Yang et al have focused their study on the prediction of the antibacterial activity by evaluating the performances of SVM classification with respect to DT and k-NN.…”

Identification of Target Associations for Polypharmacology from Analysis of Crystallographic Ligands of the Protein Data Bank

“…According to the performed ligand-based analyses, 272 and BXZ showed a high degree of similarity (MACCS fp score = 0.89; ECFP4 fp score = 0.86; TanimotoCombo = 1.68). Moreover, an analysis of the activity annotation showed that 272 is able to inhibit ERβ (IC 50 = 3.5 nM) and Hsp90 (IC 50 = 15 000 nM) with different ranges of activity values, while BXZ inhibits both ERβ (IC 50 = 3.5 nM) and Hsp90 (IC 50 = 190 nM) with submicromolar potency, based on literature data. Analysis of the predicted binding modes of BXZ and 272 into the 3BM9 (Hsp90) and 1U3Q (ERβ) (Figure S3, panel b in the Supporting Information) showed that the compounds perfectly accommodate within the investigated proteins.…”

Prediction of novel and selective TNF-alpha converting enzyme (TACE) inhibitors and characterization of correlative molecular descriptors by machine learning approaches

“…Molecular descriptors have been routinely used for quantitative description of structural and physicochemical features of molecules in QSAR and MLAs [45][46][47][48]55]. In this work, a total of 189 molecular descriptors were used, this set of descriptors was manually selected from more than 1000 descriptors described in the literature by eliminating those descriptors that are obviously redundant or irrelevant to the prediction of pharmaceutical agents [56].…”

“…On the other hand, support vector machine (SVM) and other machine learning approaches (MLA) have been consistently shown to have excellent performance for predicting various pharmacodynamic, pharmacokinetic and toxicological properties of compounds of diverse structures [45][46][47][48]. Therefore, in this study, it is of interest to test the usefulness and performance of SVM and other MLA as potential tools for the prediction of TACE inhibitors.…”